PHYSICAL REVIEW LETTERS

VOLUME 72, NUMBER 16

Measurement

of the Size Dependent Hole Spectrum

18 APRIL 1994

in CdSe Quantum

Dots

D. J. Norris, A. Sacra, C. B. Murray, and M. G. Bawendi

Massachusetts

Institute

of Technology, 77 Massachusetts
(Received

Avenue, Cambridge,

Massachusetts

02139

July 1993)

%'e combine a new synthesis with transient optical hole burning to observe and assign the size evolution (19 to 115 A diameter, cr & 5%) of a series of excited states in CdSe quantum dots. We observe an
avoided crossing around the spin orbit energy in the hole spectra for
65 A dots, indicating the importance of valence band complexities in the description of the excited states. Comparison with dc Stark
data sho~s that bleach spectra are consistent with localized carrier induced electric AeMs.

PACS numbers:

73.20.Dx, 36.40.+d, 73.20.At, 73.20. Fz

Nanometer size structures provide the opportunity to

observe and control the properties of materials as they
evolve from molecules to the bulk. In particular, semiconductor electronic transitions show interesting and
strongly size dependent properties in this size regime.
Nanometer
size semiconductor
crystallites
(quantum
dots) which are small compared to the bulk exciton Bohr
radius exhibit three dimensional quantum confinement
[1]. Quantum dots, often called "artificial atoms, possess in principle discrete electronic excitations with wave
functions delocalized within and confined by the dot
boundary. Numerous studies [1-9] have probed the lowest photoexcited state in II-VI dots. Its confinement induced shift to higher energy with decreasing diameter is
well established [I]. However, experimental investigation
of the size evolution of the higher excited states has been
diflicult. The discrete nature of these states is usually
such as distribuconcealed by sample inhomogeneities
tions in size, shape, and stoichiometry.
The observation and assignment of a series of excited
states in CdSe quantum dots, the variation of their spacings with size, and the comparison of these spacings with
recent theory are the focus of this Letter. Our ability to
observe a series of excited states depends crucially on a
recently developed synthesis which provides II-VI quan(& 5%
tum dots with a high degree of monodispersity
rms), consistent crystal structure and surface derivatiza115 A in diame12 to
tion, and control of size from
ter [9]. Unlike most samples prepared in silicate glasses
and previous samples in polymer films which typically
emit from deep trap states with low quantum yield, our
samples show "band edge" luminescence with little deep
trap emission and high quantum yield (ranging from 0. 1
to 0.9 at 10 K). The high quality of our samples is
demonstrated in Fig. 1(a) by the absorption spectra of a
series of sizes from 19 to 115 A in diameter. As many as
six transitions can be resolved in a single sample.
this sample series represents the highest
Although
quality yet reported, the main contribution to the width
of the absorption features remains sample inhomogeneities. Nanosecond pump-probe spectroscopy can further resolve the transitions [2-5]. With this technique we
optically select a subset of the quantum dot size distribu-

"

2612

tion by exciting each sample on the red edge of its first

transition with a spectrally narrow pump beam. The
pulllp-induced
change in optical density (OD) contains
both bleached and induced absorptions arising from particles already containing an electron-hole pair [Fig. 1(b)].
Comparison of pump-probe spectra [Fig. 1(b)] with absorption data [Fig. 1(a)] shows that bleached features
correspond to ground state transitions, but are more
clearly resolved. To independently verify this conclusion
we use another size-selective technique, photoluminescence excitation (PLE). PLE reveals the positions of
ground state absorptions by measuring luminescence intensity as a function of excitation energy. To measure
only a subset of the size distribution, we monitor a narrow spectral band on the blue edge of the full luminescence while scanning the excitation wavelength [3]. Our
PLE results [Figs. 1(c) and l(d)] clearly confirm that
pump-probe bleaches lie at the positions of ground state

transitions.
Pump-probe

spectra

for our entire series of dots is

shown in Fig. 2. This series spans the strong con6nement

regime in which electron-hole Coulomb interactions are
much smaller than the confinement energies. This situation is addressed in early theoretical work by combining a
simple two band isotropic effective mass model with a

spherical boundary condition and treating the electron

and hole separately [1]. The wave functions are described by "particle-in-a-sphere" envelope functions with
electron and hole angular momenta L, and Ls. However,
this approach ignores the complexities of the CdSe
valence band which is formed from Se p atomic orbitals.
The sixfold degenerate valence band is split by spin-orbit
coupling into a band-edge fourfold degenerate 1 3/2
1/2 band. For dot
band and a twofold degenerate
hole levels arising from these bands the envelope function
angular momentum, LIand the unit cell angular momentum, J, are not good quantum numbers. Only the total angular momentum, F=LI, +J, and parity are conserved. States with the same F and equal parity mix
[10,11]. Therefore, each hole level with total momentum
F (labeled as ns LF) is a mixture of Ls and Ltt + 2 envelope functions. The electron states, not aAected by the
valence band complications, are labeled as n, L, . [nest, )

0031-9007/94/72(16)/2612 (4) $06.00

1994 The American Physical Society

PHYSICAL REVIEW LETTERS

VOLUME 72, NUMBER 16

Wavelength

700

18 APRIL 1994
I

(nm)

10K

500

600

a)

OK

D(A

6
CII

3
R

b)

0QCI
I

67A

10K
i

th

C
C

10K

-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2

Energy (eV)

PLE

EO

C
62A

FIG. 2. Bleach spectra (IO K), with pump energies: 1.842,

1.984, 2.060, 2. 130, 2. 198, 2.296, 2.339, 2.366, 2.490, and
2.557 eV for 115 to 19 A dots. ISy2IS, transitions, typically

1 OK
~

1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2

bleached several percent, are set to zero energy.

Energy (eV)

FIG. I. (a) Absorption spectra (10 K) for CdSe dots embedded in optically clear poly(vinyl butyral) (PVB) films (offset for
clarity). Typical OD at the first absorption feature is 0.3-0.6.
Mean diameters are measured by TEM. Black vertical lines
mark pump frequencies. (b) Bleach spectrum, plotted as negative change in optical density, for 67 A dots (10 K). Positive
(negative) peaks are bleaches (induced absorptions).
The
hOD 0. A 10 Hz Qdashed horizontal line represents
switched Nd:YAG/dye laser system
ns pulses) produced
both pump (1.984 eV) and spectrally broad probe pulses,
separated by 12 ns. (To measure bleach lifetime, a second synchronized 10 Hz Q-switched Nd:YAG laser produced delayed
kW/cm2) was conprobe pulses. ) Pump intensity
trolled to ensure linear absorption, checked via luminescence.
Transmitted probe light was dispersed and detected with a 5 ns
gated optical multichannel analyzer. (c) Fluorescence and PLE
spectra for 67 A dots (10 K). Emission for PLE was observed
at 2.016 eY with a
5 meY bandpass. A small energy shift
(utilized in Fig. 3) exists between the pump-probe and PLE
spectra since the two techniques sample diNerent portions of the
size distribution. (d) Fluorescence and PLE spectra for
A
dots (IO K). Emission for PLE observed at 2.022 eV with a
7 me Y bandpass.

(-7

((100

-1.

-62

-0.

are counters which number hole (electron) eigenstates for

a given symmetry. l For example, the first excited state is
written as 1Sy2ISwhich contains contributions from
the following hole states: (J 3/2, Ls 0, F =3/2),
(J 3/2, Lg, 2, F 3/2), and from the split off band
(J I/2, Ls 2, F 3/2). Within the strong
confinement approximation, the Coulomb eAect is added
as a erst order perturbation. In reality, especially for the
larger dots, the Coulomb interaction further mixes these

eigenstates and alters oscillator strengths [12].

From Fig. 2 we extract the bleached absorption positions by fitting the spectra to a sum of positive(bleaches) and negative-going
(induced absorptions)
Gaussians by standard least squares methods.
Since
some spectra are difficult to interpret due to overlap of
bleaches with induced absorptions we assign only those
features which can be followed through several sizes.
Questionable peaks close to the zero crossing, such as the
broad plateau at 0.7 eV for 32 and 30 A dots, are not included. When we compare our results with theory, we
find that, as with simpler models, even the recent sophisticated eff'ective mass theory of Ref. [I I], which includes
spin-orbit coupling, valence band degeneracies, and the
of the conduction band, increasingly
nonparabolicity
overestimates the confinement energy with decreasing
size. This theory diverges in part due to its assumption of
an infinite potential barrier at the dot boundary, but also
due to the failure in general of effective mass theory for
the smallest dots, which are better described in more
molecular terms [1,11,13]. Most of the theoretical error
originates in predicting the electron level (m, ms).
Since most of the lowest allowed transitions contain the
same electron level (IS, ), plotting transition energies relative to the first excited state (ISyzlS, ) minimizes the
theoretical deviation in comparisons of theory with experimental results. Such a plot removes the common and
problematic electron contribution and yields the hole
spectrum in CdSe quantum dots.
Figure 3 (solid squares) shows our observed excited

2613

PH YSICAL R EVI EW LETTERS

VOLUME 72, NUMBER 16

18 APRIL 1994

decreasing radius

10

0.9

0.8

'

Bleach Data
PLE Data

~o

S~/21se
CO

O
~~

p7

Cl

0
2S1/P1Se-

0.6

0.5
I

Ifl

p4

1S)/21Se

II.

0.3

g 02
U

~ 2S3/21Se

1S3/21Se

0.1
M
LU

R~08$-

0.0
I

~ a

I.
1.9 2.0 2. 1 2.2 2.3 2A 2.5 2.6

..

Energy of 1S3/$1Se (eV)

FIG. 3. Excited state spacings from Fig. 2 (solid squares),

plotted relative to I S3y21$, . The x axis is the size dependent
energy of ISy21S, . Dashed lines are visual guides for our assignments.
Solid lines are predictions from Fig. 3 of Ref.
[11(b)1. PLE data from Figs. 1(c) and l(d) are plotted (open
circles) to further support the observation of an avoided crossing.

state energy spacings (relative to IS3/2IS, ) as a function

of the energy of the IS3/21S, state. We use this x axis
instead of dot diameter since the energy of the IS3/zlS,
state is experimentally more precisely measured than dot
size. Using size would introduce significant noise to the
plot due to measureinent errors. In addition, the diameters quoted in Fig. I are mean values which do not address the subset of the size distribution measured in the
bleach results. Instead, we use the energy of the IS3/21S,
state, which is easily and accurately determined and is a
better size dependent label for those dots that are actually
probed. The dashed lines in Fig. 3, visual guides for our
assignments, trace the evolution of observed transitions
with size. The solid lines are the prediction of Ref. [I 1].
We do not plot theoretical curves for all allowed transitions, but rather those which best fit the experimental results. Our assignments are based on the good agreement
between data and theory. Surprisingly, we do not observe
several transitions expected to be strongly allowed, e.g. ,
I P3/21P, ll ll. However,
these transitions, as with other
unobserved allowed transitions, poorly match the experimental points in Fig. 3, both in position and slope. These
transitions may be hidden by strong induced absorptions
in the pump-probe data.
We stress that simple eA'ective mass theory cannot account for our data. A series of states split from ISy2IS,
by spin-orbit coupling (0.42 eV) is seen in Fig. 3. These
states, which evolve into the bulk spin-orbit band (C

2614

2.0 2.2 2.4 2.6 2.8 3.0 3.2

Energy (eV)

FIG. 4. dc Stark diA'erence (field on-field olf, -5X10

V/cm internal field, 10 K) and bleach spectra for 51 A CdSe
dots. With this field and pump power close to saturation the
bleach signal is -4&& larger than the Stark signal for similar
optical densities. For the pump-probe data, if we crudely model
the dot as a delocalized electron with a "trapped, point source
hole at the dot surface, we estimate that the electric field at the
dot center is 2X IO~ V/cm. This very roughly implies onefourth of the particles are excited, consistent with the experimental data.

"

band), appear in zeroth order theory as a horizontal line

due to the similar effective masses of the bulk A and C
bands. In reality this series of states is perturbed by
nearby hole eigenstates with the same symmetry. We observe this situation in the bleach data as the avoided
crossing of the 2SI/21S, and 3SI/2IS, states for
A
dots. To confirm this claim we plot PLE data from Figs.
1(c) and 1(d) at the crossing region in Fig. 3 (open circles). The avoided crossing is intimately connected with
the mixture of Lp and Lg+2 envelope functions in the
hole eigenstates and, therefore, it is instructive to focus
on the 2SI/21S, state in Figs. 2 and 3 and follow the
change in its L/, and L/, +2 components as a function of
size. This state is a mixture of d-like (J 3/2, L/, =2,
I/2) and s-like (J=1/2, Lh =0, F=1/2) hole levels
only. All of the oscillator strength comes from the s-like
component which couples strongly to the IS, electron level. In large dots, the 2Slg21S, state has the hole mostly
in the d-like
3/2 level, and is a weak transition. After
the avoided crossing with the 3Sly21S, state, the hole
state becomes mostly s-like from the
I/2 band and
This variation in absorption
gains oscillator strength.
strength is observed in Fig. 2 and provides additional evidence for our assignments. In small dots, the hole returns
to weakly absorbing d-like character (J=3/2) after the
avoided crossing with the I S l/f21 S, state. Unfortunately,
the avoided crossing of the ISly2IS, and 2Sl/21S, states
32 A is not observed, probably due to overlap with
at
strong induced absorptions.
To understand the origin of the pump-probe spectrum,
Fig. 4 compares it with dc Stark data. The strong simiThe long
larity seems to imply a common mechanism.

-65

F=

PH YSICAL REVI EW LETTERS

VOLUME 72, NUMBER 16

lifetime of our pump-probe spectra (there is still signal

after 10 Jts) suggests that one or both of the excited carriers is localized in trap states. We argue that the electric
field produced by separation of the first electron-hole pair
causes a "Stark effect" on the absorption of a second pair
[6]. Both shallow hole surface traps, observed in band
edge luminescence studies [3,8] with a 100 ns-1 ps decay
1 ps) weakly radiative or
time, as well as longer lived
In general a
nonradiative deep trap states, contribute.
Stark field induces both a shift of state positions and a
of oscillator strength. In the small field
redistribution
limit with well separated line shapes the Stark difference
spectrum is proportional to the first derivative of the absorption spectrum, i.e., AOD(a) a: E f'(a), where E is the
electric field, is the absorption line shape, and a is energy. This result can easily be shown by expressing the line
shape as a Taylor expansion in the Stark shift and calculating the perturbing Stark interaction to second order.
However, in the small field limit with broad absorption
linewidths, larger than both the field induced shift and
states, the Stark
between interacting
the separation
difference spectrum is proportional to the second derivative of the absorption spectrum, or BOD(a) txE f"'(a)
This results from expressing the line shape in a Taylor expansion in both the Stark shift and the spacing between
interacting states. In the limit of broad linewidths, the
second derivative line shape implies that Stark bleaches
appear at the positions of ground state transitions. The
observable effect of the field is to bleach initially allowed
transitions and cause induced absorptions due to both
shifting lines and initially forbidden transitions gaining
oscillator strength. This limit is appropriate for our system in which the interacting levels are closely spaced relative to the linewidths (for example, a forbidden p-like
hole state lies in between the first two observed transitions
[11])and this description agrees with experiment. In the
Stark difference spectra, we observe bleaches at the positions of the absorption peaks, as seen previously [14, 15].
These bleaches show no field dependent shift but increase
with the field. Therefore,
in magnitude
quadratically
pump-probe bleaches in Fig. 2, which appear at the positions of ground state transitions, are consistent with a
trapped carrier induced Stark effect.
It is surprising that the first excited state bleaches are
so broad in our results. Pump-probe or PLE should considerably narrow the size distribution. The single particle
1S3/$1S, linewidth, determined from previous pumpprobe and fluorescence line narrowing experiments [2,3,
5, 7], is significantly narrower (from 4 to 16 meV, increasing with decreasing size) [7]. We believe that the combination of a sharp bleach "spike" at the pump frequency
and a broad bleach to the blue in each of our pump-probe
spectra (Fig. 2) suggests an explanation. This structure
is consistent with the existence of an absorption "tail" on
the single particle 1Sy~lS, linewidth. This tail, which
will be discussed in a separate publication, partially rein-

()

18 APRIL 1994

troduces the size distribution.

This Letter presents a measurement of the size dependence of a series of transitions in CdSe quantum dots.
Good agreement with recent theoretical predictions allows us to assign the transitions. Our data show that the
complexity of the valence band and the strong spin-orbit
coupling in CdSe dominate the size dependence of excited
state spacings. This is evident from the observation of an
avoided crossing for -65 A quantum dots, observed in
pump-probe spectra and confirmed by PLE. The strong
similarity between the full bleach spectrum and dc Stark
data suggests that electric fields produced by carrier localization induce a Stark effect in excited quantum dots.
We thank Al. L. Efros for stimulating discussions.
D.J.N. and C. B.M. benefited from NSF and NSERC fellowships, respectively. This work was funded in part by
the M IT Center for Materials Science and Engineering
(NSF-DMR-90-22933), by NSF (DMR-91-57491,
CHE-89-14953, ECS-91-18907), and by the Lucille and
David Packard Foundation.

[I] Al. L. Efros

and A. L. Efros, Fiz. Tekh. Poluprovodn. l6,

1209 (1982) [Sov. Phys. Semicond. 16, 772 (1982)]; L. E.
Brus, J. Chem. Phys. SO, 4403 (1984).
[2] A. P. Alivisatos, A. L. Harris, N. J. Levinos, M. L.
Steigerwald, and L. E. Brus, J. Chem. Phys. 89, 4001

(1988).
[3] M. G. Bawendi, W. L. Wilson, L. Rothberg, P. J. Carroll,
T. M. Jedju, M. L. Steigerwald, and L. E. Brus, Phys.
Rev. Lett. 65, 1623 (1990); M. G. Bawendi, P. J. Carroll,
W. L. Wilson, and L. E. Brus, J. Chem. Phys. 96, 946

(1992).
[4] N. Peyghambarian et al. , IEEE J. Quantum Electron. 25,
2516 (1989).
[5) U. Woggon, S. Gaponenko, W. Langbein, A. Uhrig, and
C. Klingshirn, Phys. Rev. B 47, 3684 (1983).
[6] E. F. Hilinski, P. A. Lucas, and Y. Wang, J. Chem. Phys.
$9, 3435 (1988).
[7] D. J. Norris, M. Nirmal, C. B. Murray, A. Sacra, and M.
G. Bawendi, Z. Phys. D 26, 355 (1993).
[8] M. O' Neil, J. Marohn, and G. McLendon, J. Phys. Chem.
94, 4356 (1990); A. Eychmiiller, A. Hasselbarth, L. Katsikas, and H. Weller, Ber. Bunsenges. Phys. Chem. 95, 79

(1991).
[9] C. B. Murray, D. J. Norris, and M. G. Bawendi, J. Am.
Chem. Soc. l l5, 8706 (1993).
[10] J. B. Xia, Phys. Rev. B 40, 8500 (1989).
[I I] (a) G. B. Grigoryan, E. M. Kazaryan, Al. L. Efros, and
T. V. Yazeva, Sov. Phys. Solid State 32, 1031 (1990); (b)
A. I. Ekimov et al. , J. Opt. Soc. Am. B lO, 100 (1993).
[12] Y. Z. Hu, M. Lindberg, and S. W. Koch, Phys. Rev. B
42, 1713 (1990).
[13] Y. Wang and N. Herron, Phys. Rev. B 42, 7253 (1990).
[14] F. Hache, D. Ricard, and C. Flytzanis, Appl. Phys. Lett.
55, 1504 (1989).
[15] V. L. Colvin and A. P. Alivisatos, J. Chem. Phys. 97, 730
(1992).

2615